kappa when number of events is not predetermined


I want to use kappa to assess intraobserver reliability when counting
agressions between pigs. Two observers watch videorecording of pigs
housed in pens, and they note down the time and pigs involved in each
agressive interaction, and also the intensity score.
When they both see the interaction, I just need to check if they agree
on the id number of the pigs involved, and the score. However, there
are some interactions that 1 observer may miss, o maybe the observer
who notes them down just makes them up....
In other words, what is the total number of events in this case??? I
can't build a 2*2 contingency table (observer 1 interaction seen or
not, observer 2 interactin seen or not), because there build be only 3
cells, the obs 1 not seen * obs 2 not seen!!!

However, I have seen papers using kappa for similar analyses...